Representation Learning For Speech Recognition Using Feedback Based Relevance Weighting

TL;DR

This paper introduces a novel feedback-based relevance weighting method for acoustic embedding in speech recognition, improving performance on standard datasets by adaptively selecting features from raw waveforms.

Contribution

It proposes a two-stage acoustic embedding approach with feedback-driven relevance weighting, enhancing feature selection for speech recognition.

Findings

Achieved 15% relative improvement on Aurora-4 dataset.

Achieved 7% relative improvement on CHiME-3 dataset.

Significant performance gains over baseline systems.

Abstract

In this work, we propose an acoustic embedding based approach for representation learning in speech recognition. The proposed approach involves two stages comprising of acoustic filterbank learning from raw waveform, followed by modulation filterbank learning. In each stage, a relevance weighting operation is employed that acts as a feature selection module. In particular, the relevance weighting network receives embeddings of the model outputs from the previous time instants as feedback. The proposed relevance weighting scheme allows the respective feature representations to be adaptively selected before propagation to the higher layers. The application of the proposed approach for the task of speech recognition on Aurora-4 and CHiME-3 datasets gives significant performance improvements over baseline systems on raw waveform signal as well as those based on mel representations (average relative improvement of 15% over the mel baseline on Aurora-4 dataset and 7% on CHiME-3 dataset).